Comparative analysis of graphite oxidation behaviour
based on microstructure
Heinrich Badenhorst*, Walter Focke
SARChI Chair in Carbon Materials and Technology, Department of Chemical Engineering, University
of Pretoria, Lynwood Road, Pretoria, 0002, South Africa
*Corresponding author. Tel.: +27 12 420 4989; Fax: +27 12 420 2516.
Email address: [email protected]
Abstract
Two unidentified powdered graphite samples, from a natural and a synthetic origin
respectively, were examined. These materials are intended for use in nuclear
applications, but have an unknown treatment history since they are considered
proprietary. In order to establish a baseline for comparison, the samples were compared
to two commercial flake natural graphite samples with varying impurity levels. The
samples were characterized by conventional techniques such as powder X-ray
diffraction, Raman spectroscopy and X-ray fluorescence. The results indicated that all
four samples were very similar, with low impurity levels and good crystallinity, yet they
exhibit remarkably different oxidation behaviours. The oxidized microstructures of the
materials were examined using high-resolution scanning electron microscopy at low
acceleration voltages.
The relative influence of each factor affecting the oxidation was established, enabling
a structured comparison of the different oxidative behaviours. Based on this analysis, it
was possible to account for the measured differences in oxidative reactivity. The
material with the lowest reactivity was a flake natural graphite which was characterized
as having highly visible crystalline perfection, large particles with a high aspect ratio and
no traces of catalytic activity. The second sample, which had an identical inherent
microstructure, was found to have an increased reactivity due to the presence of small
catalytic impurities. This material also exhibited a more gradual reduction in the
1
oxidation rate at higher conversion, caused by the accumulation of particles which
impede the oxidation.
The sample with the highest reactivity was found to be a milled, natural graphite
material, despite its evident crystallinity. The increased reactivity was attributable to a
smaller particle size, the presence of catalytic impurities and extensive damage to the
particle structure caused by jet milling. Despite displaying the lowest levels of crystalline
perfection, the synthetic graphite had an intermediate reactivity, comparable to that of
the highly crystalline but contaminated sample. The absence of catalytic impurities and
the needle coke-derived particle structure were found to account for this behaviour.
This work illustrates that the single most important factor when comparing unknown
graphite materials from different origins is an assessment of the oxidized microstructure.
This approach has the added benefit of identifying further potential processing steps
and limitations for material customization.
Keywords:
Microstructure,
Graphite,
Oxidation,
Impurities,
Natural,
Synthetic,
Crystallinity
1. Introduction
Graphite in its various forms is a very important industrial material, it is utilized in a
wide variety of specialized applications. These include high temperature uses where the
oxidative reactivity of graphite is very important, such as nuclear reactors. Both natural
and synthetic graphite are used in next generation high temperature reactors (VHTRs).
They comprise part of the fuel compacts for prismatic core and the fuel spheres for the
pebble bed reactors. Bulk synthetic graphite is also used extensively as a structural
component for both moderator and reflector purposes. Thus knowledge of the hightemperature oxidative behaviour of graphite is essential to understanding any
degradation which may occur due to oxidation during normal operation over the reactor
lifetime. Furthermore it is particularly important for ensuring core stability during
abnormal conditions such as air ingress into the reactor.
For these highly graphitic materials the layered structure of the ideal graphite crystal
is well established. However, in order to compare materials for a specific application the
2
number of exposed, reactive edge sites are of great importance. This active surface
area (ASA) is critical for quantifying properties like oxidative reactivity, since attack by
oxygen is only possible at exposed edges or defects [1-6]. It is evident that the ASA will
depend heavily on the manner in which crystalline regions and defects within the
material are arranged and interconnected. The concept of ASA has been around for
many decades, however due to the its low values for macrocrystalline graphite and the
complex microstructures found in these materials, it has been difficult to implement in
practice. This has led to a large number of kinetic studies reporting differing kinetic
parameters [7-12].
In addition, material morphology will be influenced by beneficiation and production
techniques like milling, which may obscure and damage the underlying structure.
Traditionally, techniques such as powder X-ray diffraction (pXRD) and Raman
spectroscopy have been used to investigate the crystalline structure of these materials.
These techniques, coupled with impurity analysis like X-ray fluorescence spectroscopy
(XRF), provide a general outline of the expected oxidation behaviour. However, they do
not allow the direct comparison of the observed oxidative reactivity for samples from
different or unknown origins.
The limited applicability of pXRD to macrocrystalline materials is well known [13]. At
a certain value the peak width is governed by the coherence of the incident beam and
not by the particle size for perfect crystals [14]. Maximum crystallite size limits of as
small as 300 nm have been suggested [15]. Raman spectroscopy on the other hand
can give ambiguous results for materials with different microstructures and defect types
since it is not possible at present to distinguish between them [16]. Thus only through
an analysis of the oxidized microstructure can the underlying crystallinity and structural
defects be identified.
Furthermore, the oxidative reactivity of graphite is very sensitive to the presence of
very low levels of impurities that are catalytically active. The catalytic activity is directly
dependent on the composition of the impurity not just the individual components. Hence
a pure metal will behave differently from a metal oxide, carbide or carbonate [17-19], a
distinction which is impossible to ascertain from XRF. In addition, the very low levels
required to significantly affect reaction rates are close to the detection limits of most
3
techniques, as such the only way to irrefutably verify the absence of catalytic activity is
through visual inspection of the oxidized microstructure.
Despite these facts, many oxidation studies do not consider the influence of, or even
simply display an image of, the microstructure of the material under investigation [912;20-28]. This makes it very difficult to compare the reactivity results obtained by these
studies on a like-for-like basis. The aim of this investigation is to demonstrate that it is
possible to discern between unknown samples for the purposes of materials selection,
using only the most simplistic kinetic test and an investigation of the oxidized
microstructure.
Two unidentified powdered graphite samples, from a natural and a synthetic origin
respectively, were examined. These materials are intended for use in nuclear
applications, but have an unknown treatment history since they are considered
proprietary. In order to establish a baseline for comparison, the samples were compared
to two commercial flake natural graphite samples with varying impurity levels. The
natural graphite flakes are known to be highly crystalline and represent a cheap
archetypal material with a simple structure.
The samples were characterized for any discernable differences using conventional
methods, such as pXRD, Raman spectroscopy and XRF. The samples were then
partially oxidized and visually inspected using high-resolution scanning electron
microscopy (SEM). Since the kinetic parameters were not desired, a quick and simple
oxidation test was done, namely a single temperature ramp at a fixed rate in pure
oxygen. The objective is to distinguish between the factors that influence the oxidative
reactivity for each sample across different segments of the conversion range. The
relative influence of each factor affecting the oxidation was established, enabling a
structured comparison of the different oxidative behaviours. This work illustrates that the
single most important factor when comparing unknown graphite materials from different
origins is an assessment of the oxidized microstructure. This approach has the added
benefit of identifying further potential processing steps and limitations for material
customization.
4
2. Materials and Methods
Four powdered graphite samples were examined during the investigation. Powders
with relatively small particle sizes were chosen to eliminate bulk effects such as porosity
and to ensure the absence of any mass transfer limitations. In addition, a high purge
rate (500 ml.min-1), a small sample size (~2 mg) and the use of a flat sample pan were
used as a consequence of extensive investigation [29]. The absence of transfer
limitations was confirmed by control experiments, during which the purge gas flow rate
was varied while the same oxidation behaviour and rate were observed.
The first two samples were proprietary nuclear-grade graphite samples, one from a
natural source (NNG) and one synthetically produced material (NSG). Both samples
were intended for use in the nuclear industry and were probably subjected to high levels
of purification. The ash content of these samples was very low, with the carbon content
being >99.8 mass % determined by complete oxidation at 1000 ºC. However, the exact
histories of both materials are not known.
The third graphite sample (RFL grade) was obtained from a commercial source
(Graphit Kropfmühl AG, Germany). This is a large-flake, natural graphite powder and
was purified by the supplier with an acid treatment and an alkali roasting step using
sodium carbonate [30] to a reported purity of >99.9 mass %. The fourth sample was
produced by heating the RFL sample to 2750 ºC for 6 hours in a TTI furnace (Model:
1000-2560-FP20). This sample was designated PRFL since the treatment was
expected to further purify the material. Since the material is a naturally occurring mineral
it is fluid deposited under high pressure and temperature conditions during the creation
of metamorphosed siliceous or calcareous sediments [31]. Thus the treatment is not
expected to significantly modify the microstructure.
All
thermal
oxidation
was
conducted
in
a
TA
Instruments
SDT
Q600
thermogravimetric analyzer (TGA). As a comparative indication of reactivity, the
samples were subjected to oxidation in pure oxygen (instrument grade) at atmospheric
pressure under a temperature programme of 4 ºC/min in the TGA up to 1000 ºC and
100% conversion. On the other hand, for examination the samples were all oxidized to a
burn-off of around 30%, at which point the oxidizing atmosphere was rapidly changed to
inert. SEM images were obtained using an ultra-high-resolution field-emission
5
microscope (Zeiss Ultra Plus 55 FEGSEM) equipped with an in-lens detection system
operating at an acceleration voltage of 1 kV. A working distance of between 2 and 3 mm
was used and the powders were lightly deposited on carbon tape without any additional
sample preparation.
The pXRD spectra of the graphite samples were obtained on a PANalytical X-pert
Pro powder diffractometer with variable divergence and receiving slits and an
X’celerator detector using iron-filtered cobalt Ka radiation. The Raman spectra of the
samples were obtained using a Dilor XY Raman spectrometer using the λ = 514.5 nm
laser line of a coherent Innova 90 Ar+ laser.
The impurity constituents of the samples were analysed using an ARL9400 XP+
Sequential XRF analyser and Uniquant software. The analysis was done for all
elements between Na and U, but only elements found above the detection limits are
reported. The carbon content is calculated by difference.
2.1 Analytical results
The dimensionless degree of conversion may be defined as:
a=
m0 - m
m0
(1)
Where m is the sample mass at any time and m0 is the initial sample mass. Thus the
reaction rate is given by:
da
1 dm
=dt
m0 dt
(2)
The measured oxidation rate as a function of temperature is shown in Figure 1.
6
Fig. 1: Reactivity comparison (A) degree of conversion and (B) reaction rate
In order to quantify the observed behaviours it is convenient to define certain
characteristic aspects of the experimentally obtained curves. This classification allows a
systematic comparison of the differences between the materials and is not intended as
a comprehensive analysis of the oxidation behaviour. Differences in the measured
characteristics can then be linked to the microstructure. A representative reaction rate
curve is given in Figure 2 (A) with relevant annotations.
Fig. 2: Standard reaction rate (A) and conversion (B) curves
The onset temperature may be defined as the x-intercept of a linear fit to the steady
region of oxidation rate increase following the initial induction period. The NNG sample
is unique in the sense that it displays two comparatively steady regions, which are
visible in Figure 3 (A).
7
Fig. 3: Reaction rate (A) and conversion (B) curves for NNG
Thus it is necessary to define the onset temperature as the x-intercept of a linear fit
in the first steady region of oxidation rate increase. The steady gradient, however,
should be chosen as the slope in the region where the majority of the conversion has
taken place, as shown in Figure 2 (B). This is shown to be the second steady region for
the NNG sample as indicated by the conversion curve in Figure 3 (B). The final
parameter of importance is the decline gradient, which is an indication of the rate at
which oxidation decreases to zero after the peak rate has been achieved. These
parameters were calculated for the samples under consideration and are shown in
Table 1.
Table 1: Oxidation parameters
Sample
Onset
Steady Gradient
-1
-1
Decline Gradient
Temp (ºC)
(s .ºC )
(s-1.ºC-1)
NNG
572
0.70E-04
-1.70E-03
RFL
696
4.54E-04
-0.66E-03
NSG
704
5.10E-04
-1.58E-03
PRFL
760
3.78E-04
-1.29E-03
8
Fig. 4: pXRD spectra
The pXRD spectrum for each sample is shown in Figure 4. The spectra are very
similar, indicating that the four samples are highly crystalline as expected. The only
exception is NNG, which shows the presence of the rhombohedral stacking fault. This
form of graphite can be easily identified by its (100) and (110) reflections [32] at 2q =
50.9a and 54.4a (CoKa) respectively, as indicated by the arrows. The Raman spectra
for all the samples are shown in Figure 5.
Fig. 5: Raman spectra
All four samples also have very similar Raman spectra. The only irregularity is the
PRFL sample, which has no visible D peak at a Raman shift of 1360 cm-1 [16]. This
9
would indicate that the sample is remarkably crystalline, with virtually no disorder. That
being said, the D peaks of all the samples were relatively small, indicating low disorder
in all of them.
Table 2: XRF compositional analysis (Mass %, B.D.L. = Below detection limits)
NNG
RFL
NSG
PRFL
Si
0.01
0.01
0.01
B.D.L.
Fe
B.D.L.
0.01
B.D.L.
B.D.L.
Mg
0.02
0.03
0.01
0.03
Na
0.10
0.08
0.02
0.02
S
0.06
0.02
0.01
0.01
Mo
0.01
B.D.L.
0.01
B.D.L.
Th
B.D.L.
B.D.L.
0.01
B.D.L.
Carbon
99.81
99.85
99.95
99.95
The XRF analysis of the samples are given in Table 2, indicating the impurities
present. The results confirm that all the samples have very low impurity levels, less than
2000 ppm in all cases. In the context of the mass loss curves given in Figure 1 (A) such
low amounts are not readily evident. In addition, at these low levels the absolute
accuracy is disputable and the possibility exists that the contamination may have
occurred due to storage and handling before analysis.
Despite the wide variation in oxidation behaviours demonstrated by the samples, the
conventional characterization techniques indicate that the samples are very similar. The
minor observed differences are clearly not sufficient for materials comparison or
selection.
2.2 Microstructural Investigation
This section only highlights key aspects of the partially oxidized microstructures.
This is followed by a detailed discussion to link the observed microstructures to the
measured oxidation behaviours.
10
2.2.1 Crystallinity
The microstructure of a sample may be subdivided into two categories, crystallinity
and particle geometry. The layered nature of all four samples can easily be discerned
as demonstrated in Figure 6.
Fig. 6: Samples edge-on: (A) NNG, (B) RFL, (C) NSG, (D) PRFL
This makes it easy to identify the basal plane of each sample. A representative
portion of the basal plane for each material is shown in Figure 7, demonstrating the
smooth relatively intact basal surfaces of all but the NSG sample.
Fig. 7: Sample basal planes: (A) NNG, (B) RFL, (C) NSG, (D) PRFL
In order to understand the crystallinity of the material it is also important to examine
the edges from above as in Figure 8.
Fig. 8: Sample edges: (A) NNG, (B) RFL, (C) NSG, (D) PRFL
The edges are disordered and erratic for all but the PRFL graphite.
11
2.2.2 Particle geometry
This aspect of the microstructure includes both particle size and shape or structure.
The relative differences in particle size are shown in Figure 9, where all four images
were taken at the same magnification (500x).
Fig. 9: Particle sizes: (A) NNG, (B) RFL, (C) NSG, (D) PRFL
The particles have very different shapes and structures, which is not unexpected
considering their differing origins and unknown history. Nonetheless it is clear that the
NNG and NSG samples have significantly smaller particles. A representative particle for
each material is shown in Figure 10.
Fig. 10: Particle shapes: (A) NNG, (B) RFL, (C) NSG, (D) PRFL
As expected the flat, flake-like natures of the RFL and PRFL natural graphite
samples are readily visible. However, the nuclear-grade natural graphite, NNG, has a
very complex, highly disordered structure. The NSG particles are anisotropic, having a
very irregular and intricate structure along the A direction but a fairly smooth, continuous
surface in the B direction. Aside from the inherent microstructure, impurities may also
influence the oxidized material structure.
2.2.3 Catalytic impurities
As mentioned earlier, the absolute accuracy of the XRF compositional analysis is not
irrefutable. The only way to be certain of the presence or absence of catalytically active
12
particles is to examine the particle surfaces at very high magnification. At high
acceleration voltages (>5 kV) electrons penetrate deep into the material, leading to a
loss in surface resolution. Thus low acceleration voltages must be used to ensure that
surface detail is captured with sufficient sensitivity to resolve very small particles. When
the NNG and RFL samples are closely examined, the presence of catalytic impurities
was found, as shown in Figure 11.
Fig. 11: Catalytically active particles
The particles trace random, erratic channels in the graphite. The catalyst particles
tend to unzip the graphite, leading to large furrows being formed as uncatalyzed
oxidation proceeds along the channel walls. These signatures are generally easy to
find, allowing the detection of extremely low levels of catalyst activity. A variety of
catalytic behaviours were observed, as shown in Figure 12 (A) to (C), along with a few
instances of pitting catalysts evident in Figure 12 (D).
Fig. 12: Catalytic behaviours: (A) Random, (B) Preferred direction, (C) Wetting, (D)
Pitting
No catalytic activity was observed in either the NSG or PRFL sample despite
extensive examination.
13
2.2.4 Inhibiting impurities
Particles that are not catalytically active will remain stationary, and as they shrink
during oxidation, they will accumulate at the edges. An example of such an
accumulation along an edge is shown in Figure 13 (A).
Fig. 13: Inhibiting impurities: (A) Initial accumulation, (B) Oxidation inhibition, (C)
Formation of characteristic saw-tooth edge
In some cases these particles will accumulate into ridges along the edge, which are
also visible in Figure 13 (A). If the inactive particle is large enough or oxidation of the
surrounding region has proceeded to large extent, the particle will tend to shield a layer
of graphite from oxidation. This layer will shield subsequent layers, leading to a cascade
effect and the formation of a triangular inhibition ridge, as seen in Figure 13 (B). On a
macro scale this phenomenon will impact the overall oxidation rate if sufficient impurity
particles are present and will lead to the formation of characteristic saw-tooth edges,
visible in Figure 13 (C). These edges were only observed in the RFL graphite and were
found to be relatively widespread when the sample was oxidized to a higher level of
burn-off.
3. Discussion
Based on the onset temperatures given in Table 1, the initial reactivity of the
samples may be ranked as: PRFL < NSG < RFL < NNG. After an initial rapid increase in
the oxidation rate, the NNG sample does, however, have a comparatively low oxidation
rate during the steady region, where the majority of the conversion takes place. Of the
three samples, the PRFL sample has the lowest steady oxidation gradient. This implies
that as the temperature is increased, this sample has the lowest incremental rise in
14
reaction rate. Lastly, the only sample to show a considerably lower decline gradient is
the RFL material, signifying that the decay in the reaction rate was slower than that of
any of the other samples.
Before the microstructures are considered, it is important to assess the presence
and influence of the catalytically active impurities. Catalysts of varying activity were
identified in the NNG and RFL samples. These particles tend to disrupt the flat and
uniform surfaces of the oxidized graphite, as they trace random, irregular channels
throughout the structure. This is most apparent when the edges are viewed from above
as in Figure 8. The basal plane also shows traces of their activity in the form of small
erratic ridges. However, only in the case of pitting catalysts is there gross structural
disruption of the basal plane – other catalysts simply move across the surface without
much damage, as seen in Figure 7. When viewed edge-on the random undulations
caused by the catalysts are apparent; however, the surfaces remain fairly smooth in the
vertical direction, across several microns in some cases as is visible in Figure 6. This
indicates that oxidation is proceeding homogeneously along these faces.
The smooth oxidation of edge faces is most apparent in the PRFL sample, where flat
edge surfaces measure well over ten microns in height. The excellent crystallinity of the
material is noticeable when the edges are viewed from above, demonstrating the clean
120° angles expected for pristine graphite. Despite the catalytic activity, all three natural
graphite samples exhibit very high levels of crystallinity as expected, demonstrated by
the intact basal plane and smooth edges. The synthetic graphite, on the other hand,
shows a highly pocked and roughened surface. This indicates the presence of high
levels of crystalline defects. The presence of these defects is also apparent when the
edges are examined. The synthetic material has very narrow edge segments across
which continuous oxidized surfaces are apparent.
The NNG and NSG materials have significantly smaller particle sizes than the large
flat flakes of the RFL and PRFL samples, shown in Figure 9. The flake structures of the
RFL and PRFL samples have a very high aspect ratio, with the flake diameter being an
order of magnitude more than the flake thickness. Despite the catalytic disruption, the
large, flat flake character of the RFL graphite is still visible and is similar to that of the
PRFL sample, as demonstrated in Figure 10 (B) and (D). This is not unexpected since
15
the heat treatment was not likely to modify the microstructure of the material. The NNG
particle shown in Figure 10 (A), however, is a complex jumble of what appears to be
highly agglomerated small flakes. The reason for this shape is understood when the asreceived material in Figure 14 below is examined.
Fig. 14: "Potato-shaped" NNG particle
The particles are a result of extensive damage caused by jet milling of the sample to
produce so-called "potato shaped" graphite. The milling causes the highly malleable
natural graphite flakes to bend, buckle and fold over themselves, forming a structure
similar to a ball of crumpled paper. The original particles probably had a flake-like
structure similar to the RFL material. When oxidized, the inner complex structure of the
particle is revealed. A complex structure is also found when the synthetic graphite
particles are examined. In Figure 9 (C), a collection of long, thin particles is visible. One
such particle is viewed at an oblique angle in Figure 10 (C). These characteristic
particles are derived from the needle coke filler commonly used [33] in nuclear-grade
graphite. The particles tend to be highly oriented along the long axis and, as can be
seen in Figure 10 (C), they have a very complex and irregular internal structure.
Nonetheless, despite having a highly defective structure, the basal orientation of the
outer surface is easily discernable and the layered nature of the underlying material is
readily visible, as can be seen in Figure 6 (C). Thus attack is mainly possible from the
edges exposed at the short axes and the needle shape of the particle shape remains
consistent up to high conversions.
It is now possible to assess the experimentally measured oxidation behaviours in
terms of the observed microstructural features. The low oxidative reactivity of the PRFL
16
sample is not surprising given its lack of catalytically active impurities, large particle
size, high edge-to-basal aspect ratio and excellent crystallinity. The steady gradient of
the oxidation rate should always be considered in combination with the onset
temperature. A sample may have a lower steady gradient, but if the onset occurs at a
much lower temperature, the sample will have a significantly increased reaction rate
when isothermally oxidized at the onset temperature of a more stable sample. This
makes the oxidative stability of the PRFL material even more remarkable since it shows
the lowest steady gradient but at a much higher onset temperature than the other
samples.
The exceptional significance of very low levels of catalytically active impurities is
demonstrated by the RFL sample. Microstructurally the material is identical to the PRFL
material, but the onset temperature has been significantly reduced and the steady
gradient of the reaction rate has increased. This is due to the nature of the catalytic
activity, which unzips the graphite, creating large amounts of additional surface area
where uncatalyzed oxidation takes place. So much so, that the material has a lower
onset temperature than the less ideal synthetic graphite, but a slightly lower steady
gradient.
The synthetic graphite has the highest number of crystalline defects and a small
particle size. However, the shape of the needle coke particles tends to shield the
internal structure and slow oxidation. Regardless, the complex microstructure with
multiple exposed edges and defects leads to an increase in the steady oxidation
gradient and a drop in the onset temperature relative to PRFL, despite the absence of
catalytic activity. It may be expected that, if the material contained significant amounts
of less structured synthetic materials such as particles derived from binder pitch or other
coke varieties, the steady gradient would be further increased and the onset
temperature reduced.
Despite the good crystallinity demonstrated by the NNG sample, as expected for a
natural graphite, the reduced particle size, catalytic impurities and extensive damage
caused by jet milling resulted in the sample with the lowest onset temperature. The
steady gradient is the lowest of all the samples but at significantly lower temperatures.
The sample also exhibits a unique oxidation curve with a dramatically reduced gradient
17
after the initial gradient. This behaviour may be explained based on the observed
microstructure.
In tracing random pathways through the graphite, catalyst particles will tend to
collide and coalesce, reducing their activity. Since the NNG particles are significantly
reduced in size and severely damaged, it is plausible that this will occur much earlier
compared to the RFL sample, for example. Secondly, the material is highly crystalline
with an original structure probably similar to the RFL flakes. Thus it seems reasonable
to conclude that the initial onset and rapid increase in oxidation rate is caused by the
catalyst particles. After a short induction period these particles and their channels
coalesce, resulting in reduced activity and the attainment of a pseudo steady state in
terms of additional surface area creation. At this point the inherent microstructure of the
particle controls the reaction. Given the crystallinity and flake-like nature of the parent
particles, it is then not unexpected to see a dramatic drop in the oxidation rate gradient.
Thus if the catalysts are removed and onset occurs at a high temperature, it would not
be unexpected to observe a steady gradient similar to those of the RFL or NSG
samples. The extensive damage to the particle and the additional active surface area
created by the jet milling does, however, mean that even in the absence of any catalyst
activity the NNG material will not achieve the stability of the PRFL sample.
Finally, the reduced decline gradient found for the RFL material compared to the
other samples may also be explained in terms of the observed microstructure.
Impurities which are not catalytically active tend to accumulate at the particle edges as
the particle is eroded away during oxidation. Towards high conversions, these particles
may reach significant concentrations at the edges. Since the particles do not participate
in oxidation they tend to shield a small area of the graphite, which leads to the formation
of characteristic saw-tooth edges. These edges were found to be widespread in the RFL
graphite, which explains the more gradual decline in the oxidation rate, as the surface
area available for reaction is progressively reduced towards high conversions.
5. Conclusions and recommendations
The oxidative reactivity of graphite is important for a wide variety of applications.
Kinetic investigations seldom relate the observed oxidation behaviour to the material
18
microstructure, which makes sensible comparisons difficult. The aim of this investigation
is to demonstrate that it is possible to establish a link between the experimentally
determined burn-off characteristics of different, unknwon graphite materials and their
observed oxidized microstructures. Four powdered graphite samples were examined,
two of which were from natural and synthetic origins respectively and intended for
nuclear applications. These were compared to two flake natural graphite materials with
different impurity levels.
The samples were characterized using conventional methods, namely ash content,
pXRD, Raman spectroscopy and XRF. The results were very similar and confirmed that
all four were samples of very high purity and good crystallinity. The materials were then
oxidized and visually inspected using high-resolution scanning electron microscopy
(SEM). The experimentally determined burn-off behaviour was quantified by selecting
three key features of the oxidation curves: onset temperature, steady oxidation rate
gradient and decline gradient of oxidation rate.
The PRFL sample exhibited the lowest reactivity of all the samples, based on the
onset temperature. This was directly linked to the absence of any catalytically active
impurities, the excellent crystallinity of the material and the flake particle geometry with
a large size and high aspect ratio. The PRFL material is simply a purified version of the
RFL material, and as such they have identical microstructures. However, the RFL
material showed a significant increase in oxidative reactivity based on onset
temperature and oxidation gradient. A variety of catalytically active particles were
discovered in this sample, which accounted for the increased reactivity. In addition, this
sample exhibited the only perceptibly lower decline gradient. This more gradual decline
in the oxidation rate was found to be caused by inhibiting impurities which accumulate
at the particle edges towards high conversions.
The sample with the highest reactivity was found to be one of the natural graphite
material despite the good crystallinity that it demonstrated. The increased reactivity was
attributable to a smaller particle size, the presence of catalytic impurities and extensive
damage to the particle structure caused by jet milling. This material exhibited a slightly
more complex burn-off behaviour due to the interplay of crystallinity, particle geometry
and catalytic activity. Despite displaying the lowest levels of crystalline perfection, the
19
synthetic graphite had an intermediate reactivity similar to that of the RFL material. The
absence of catalytic impurities and the needle coke-derived particle structure were
found to account for this behaviour.
Thus it is possible to link the key aspects of the experimentally determined oxidation
behaviours to the observed microstructures. This enables like-for-like comparison of
materials for selection based on the application. For example, on the basis of this
investigation, it is clear that the oxidative reactivity of the NNG sample may be improved
by purification, but due to the damaged structure it cannot achieve the stability observed
for the PRFL material, despite both being natural graphite samples. Furthermore,
despite having similar reactivities, the NSG and RFL materials have vastly different
microstructures and therefore would not be equally suitable for applications where, for
example, surface area is very important. In conclusion, it is not enough to characterize
graphite from different sources using conventional techniques alone for the purposes of
material comparison. It is recommended that kinetic parameters for the oxidative
reactivity of graphite be reported only when accompanied with a detailed microstructural
analysis.
Acknowledgments
This work is based on research supported by the Skye Foundation and the South
African Research Chairs Initiative of the Department of Science and Technology (DST)
and the National Research Foundation (NRF). Any opinions, findings and conclusions
or recommendations expressed in this report are those of the authors and therefore
Skye, the NRF and the DST do not accept any liability with regard thereto.
6. References
[1] Laine NR, Vastola FJ, Walker PL. Importance of active surface area in the carbonoxygen reaction. Journal of Physical Chemistry. 1963;67:2030-4.
[2] Thomas JM. Topographical studies of oxidized graphite surfaces: a summary of the
present position. Carbon. 1969;7:350-64.
20
[3] Bansal RC, Vastola FJ, Walker PL. Studies on ultra-clean carbon surfaces - III.
Kinetics od chemisorption of hydrogen on graphon. Carbon. 1971;9:185-92.
[4] Radovic LR, Walker PL, R.G. J. Importance of carbon active sites in the gasification
of coal chars. Fuel. 1983;62:849-56.
[5] Walker PL, R.L. J, J.M. T. An update on the carbon-oxygen reaction. Carbon.
1991;29:411-21.
[6] Arenillas A, Rubiera F, Pevida C, Ania CO, Pis JJ. Relationship between structure
and reactivity of carbonaceous materials. Journal of Thermal Analysis and
Calorimetry. 2004;76:593-602.
[7] Contescu CI, Guldan T, Wang P, Burchell TD. The effect of microstructure on air
oxidation resistance of nuclear graphite. Carbon. 2012;50(9):3354-66.
[8] Hahn JR. Kinetic study of graphite oxidation along two lattice directions. Carbon.
2005;43:1506-11.
[9] Guo W, Xiao H, Guo W. Modelling of TG curves of isothermal oxidation of graphite.
Mater Sci Eng 2008;474:197–200.
[10] Kim ES, Lee KW, No HC. Analysis of geometrical effects on graphite oxidation
through measurement of internal surface area. Journal of Nuclear Materials.
2006;348:174-80.
[11] Xiaowei L, Jean-Charles R, Suyuan Y. Effect of temperature on graphite oxidation
behaviour. Nucl Eng Des 2004;227:273–80.
[12] Moormann R, Hinssen H-K, Kühn K. Oxidation behaviour of an HTR fuel element
matrix graphite in oxygen compared to a standard nuclear graphite. Nucl Eng Des
2004;227:281–4.
[13] Klug HP, Alexander LE. X-ray diffraction procedures for polycrystalline and
amorphous materials. New York: Wiley; 1954.
[14] Dinnebier RE, Billinge SJL. Powder diffraction: theory and practice. Cambridge:
Royal Society of Chemistry; 2008.
[15] Alexander L. The synthesis of X-Ray spectrometer line profiles with application to
crystallite size measurements. Journal of Applied Physics. 1954;25(2):155-61.
21
[16] Ferrari AC. Raman spectroscopy of graphene and graphite: Disorder, electron–
phonon coupling, doping and nonadiabatic effects. Solid State Commun
2007;143(1–2):47–57.
[17] Baker RTK. Factors controlling the mode by which a catalyst operates in the
graphite-oxygen reaction. Carbon. 1986;24:715-7.
[18] Yang RT, Wong C. Catalysis of carbon oxidation by transition metal carbides and
oxides. Journal of Catalysis. 1984;85:154-68.
[19] McKee DW, Chatterji D. The catalytic behaviour of alkali metal carbonates and
oxides in graphite oxidation reactions. Carbon. 1975;13:381-90.
[20] Penner SS, Richards MB. Oxidation of nuclear-reactor-grade graphite. Energy
1988;13:461–8.
[21] Li L, Tan Z-C, Meng S-H, Wang S-D, Wu D-Y. Kinetic study of the accelerating
effect of coal-burning additives on the combustion of graphite. J Therm Anal
Calorim 2000;62:681–5.
[22] Jiang W, Nadeau G, Zaghib K, Kinoshita K. Thermal analysis of the oxidation of
natural graphite – the effect of particle size. Thermochim Acta 2000;351:85–93.
[23] Zaghib K, Song X, Kinoshita K. Thermal analysis of the oxidation of natural
graphite: isothermal kinetic studies. Thermochim Acta 2001;371:57–64.
[24] Cebulak S, Smieja-Krol B, Duber S, Misz M, Morawski AW. Oxyreactive thermal
analysis a good tool for the investigation of carbon materials. J Therm Anal
Calorim 2004;77:201–6.
[25] Hinssen H-K, Kühn K, Moormann R, Schlögl B, Fechter M, Mitchell M. Oxidation
experiments and theoretical examinations on graphite materials relevant for the
PBMR. Nucl Eng Des 2008;238(11):3018–25.
[26] Contescu CI, Azad S, Miller D, Lance MJ, Baker FS, Burchell TS. Practical
aspects for characterizing air oxidation of graphite. J Nucl Mater 2008;381:15–24.
[27] Yu X, Brissonneau L, Bourdeloie C, Yu S. The modeling of graphite oxidation
behavior for HTGR fuel coolant channels under normal operating conditions. Nucl
Eng Des 2008;238(9):2230–8.
22
[28] El-Genk MS, Tournier JP. Comparison of oxidation model predictions with
gasification data of IG-110, IG-430 and NBG-25 nuclear graphite. J Nucl Mater
2012;420:141–58.
[29] Badenhorst H, Rand B, Focke W. A generalized solid state kinetic expression for
reaction interface-controlled reactivity. Thermochimica Acta. 2013;562:1-10.
[30] Lu XJ, Forssberg E. Preparation of high-purity and low-sulphur graphite from
Woxna fine graphite concentrate by alkali roasting. Minerals Engineering.
2002;15:755-7.
[31] Luque FJ, Pasteris JD, Wopenka B, Rodas M, Barranechea JF. Natural fluiddeposited graphite: mineralogical characteristics and mechanisms of formation.
American Journal of Science. 1998;298:471-98
[32] Parthasarathy G, Sreedhar B, Chetty TRK. Spectroscopic and X-ray diffraction
studies on fluid deposited rhombohedral graphite from the eastern Ghats Mobile
Belt, India. Curr Sci 2006;90:995–1000.
[33] Nightingale RE. Nuclear Graphite. New York: Academic Press; 1962.
23